ollama remote adapter works

llama_toolchain/cli/api/api.py

@@ -10,6 +10,7 @@ from llama_toolchain.cli.subcommand import Subcommand
 
 from .build import ApiBuild
 from .configure import ApiConfigure
+from .start import ApiStart
 
 
 class ApiParser(Subcommand):
@@ -26,3 +27,4 @@ class ApiParser(Subcommand):
         # Add sub-commands
         ApiBuild.create(subparsers)
         ApiConfigure.create(subparsers)
+        ApiStart.create(subparsers)

llama_toolchain/cli/api/build.py

@@ -7,9 +7,6 @@
 import argparse
 import json
 import os
-import random
-import string
-import uuid
 from pydantic import BaseModel
 from datetime import datetime
 from enum import Enum
@@ -25,10 +22,6 @@ from llama_toolchain.common.config_dirs import BUILDS_BASE_DIR
 from llama_toolchain.distribution.datatypes import *  # noqa: F403
 
 
-def random_string():
-    return "".join(random.choices(string.ascii_letters + string.digits, k=8))
-
-
 class BuildType(Enum):
     container = "container"
     conda_env = "conda_env"
@@ -42,6 +35,8 @@ class Dependencies(BaseModel):
 def get_dependencies(
     provider: ProviderSpec, dependencies: Dict[str, ProviderSpec]
 ) -> Dependencies:
+    from llama_toolchain.distribution.distribution import SERVER_DEPENDENCIES
+
     def _deps(provider: ProviderSpec) -> Tuple[List[str], Optional[str]]:
         if isinstance(provider, InlineProviderSpec):
             return provider.pip_packages, provider.docker_image
@@ -60,7 +55,9 @@ def get_dependencies(
 
         pip_packages.extend(dep_pip_packages)
 
-    return Dependencies(docker_image=docker_image, pip_packages=pip_packages)
+    return Dependencies(
+        docker_image=docker_image, pip_packages=pip_packages + SERVER_DEPENDENCIES
+    )
 
 
 def parse_dependencies(
@@ -88,7 +85,6 @@ def parse_dependencies(
 
 
 class ApiBuild(Subcommand):
-
     def __init__(self, subparsers: argparse._SubParsersAction):
         super().__init__()
         self.parser = subparsers.add_parser(
@@ -125,8 +121,8 @@ class ApiBuild(Subcommand):
         self.parser.add_argument(
             "--name",
             type=str,
-            help="Name of the build target (image, conda env). Defaults to a random UUID",
-            required=False,
+            help="Name of the build target (image, conda env)",
+            required=True,
         )
         self.parser.add_argument(
             "--type",
@@ -153,11 +149,10 @@ class ApiBuild(Subcommand):
             )
             return
 
-        name = args.name or random_string()
         if args.type == BuildType.container.value:
-            package_name = f"image-{args.provider}-{name}"
+            package_name = f"image-{args.provider}-{args.name}"
         else:
-            package_name = f"env-{args.provider}-{name}"
+            package_name = f"env-{args.provider}-{args.name}"
         package_name = package_name.replace("::", "-")
 
         build_dir = BUILDS_BASE_DIR / args.api
@@ -176,7 +171,7 @@ class ApiBuild(Subcommand):
         }
         with open(package_file, "w") as f:
             c = PackageConfig(
-                built_at=str(datetime.now()),
+                built_at=datetime.now(),
                 package_name=package_name,
                 docker_image=(
                     package_name if args.type == BuildType.container.value else None

llama_toolchain/cli/api/start.py

@@ -0,0 +1,83 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+import argparse
+
+from pathlib import Path
+
+import pkg_resources
+import yaml
+
+from llama_toolchain.cli.subcommand import Subcommand
+from llama_toolchain.distribution.datatypes import *  # noqa: F403
+
+
+class ApiStart(Subcommand):
+    def __init__(self, subparsers: argparse._SubParsersAction):
+        super().__init__()
+        self.parser = subparsers.add_parser(
+            "start",
+            prog="llama api start",
+            description="""start the server for a Llama API provider. You should have already built and configured the provider.""",
+            formatter_class=argparse.RawTextHelpFormatter,
+        )
+        self._add_arguments()
+        self.parser.set_defaults(func=self._run_api_start_cmd)
+
+    def _add_arguments(self):
+        self.parser.add_argument(
+            "--yaml-config",
+            type=str,
+            help="Yaml config containing the API build configuration",
+            required=True,
+        )
+        self.parser.add_argument(
+            "--port",
+            type=int,
+            help="Port to run the server on. Defaults to 5000",
+            default=5000,
+        )
+        self.parser.add_argument(
+            "--disable-ipv6",
+            action="store_true",
+            help="Disable IPv6 support",
+            default=False,
+        )
+
+    def _run_api_start_cmd(self, args: argparse.Namespace) -> None:
+        from llama_toolchain.common.exec import run_with_pty
+
+        config_file = Path(args.yaml_config)
+        if not config_file.exists():
+            self.parser.error(
+                f"Could not find {config_file}. Please run `llama api build` first"
+            )
+            return
+
+        with open(config_file, "r") as f:
+            config = PackageConfig(**yaml.safe_load(f))
+
+        if config.docker_image:
+            script = pkg_resources.resource_filename(
+                "llama_toolchain",
+                "distribution/start_container.sh",
+            )
+            run_args = [script, config.docker_image]
+        else:
+            script = pkg_resources.resource_filename(
+                "llama_toolchain",
+                "distribution/start_conda_env.sh",
+            )
+            run_args = [
+                script,
+                config.conda_env,
+            ]
+
+        run_args.extend(["--yaml_config", str(config_file), "--port", str(args.port)])
+        if args.disable_ipv6:
+            run_args.append("--disable-ipv6")
+
+        run_with_pty(run_args)